Upscaling is the process of taking a small image and enlarging it to display on a high-resolution display. When you watch a DVD on your High-Definition television, the image is being upscaled to fit the screen.
The opposite of upscaling is downscaling. This would happen if you watched a HD Blu-ray movie on a standard-definition television. When you make an image smaller, it will continue to look as good as it possibly can, but when you enlarge an image, it can suffer a loss in visual fidelity. That's where upscaling becomes important.
When dealing with consumer-level home theatre equipment, there are three common types of upscaling used. These are:
Each method constitutes a different mathematical algorithm, and each one processes the same image differently. It's easiest to illustrate these differences with images. Our example city scene here is very small, and we want to double it in size. We'll use all the major upscaling methods and show you what effect they have.
In this example, all of the pixels from the original image have been spread out to fill a space double the size. As you can see, the image is still there, but three quarters of the area is empty, black space - without any image data at all.
You're not likely to see this done in practice, but it gives you a good idea of the 'blanks' that upsclaing has to fill in: quite a lot!
Nearest Neighbour is the simplest, fastest and least visually appealing method available. You generally won't see this method used in home theatre equipment, for obvious reasons. The pixel structure of the image has become very apparent, and you'd have to sit twice as far away from the screen as normal to make it look smooth again.
This process simply doubles the size of the pixels from 1x1 to 2x2 to fill in the gaps.
Bilinear Interpolation is the next easiest method to implement. It scales the image up smoothly (most of the time), and interpolates 'missing' pixels by averaging them against their neighbours. It looks at pixels in groups of 4 (2x2) in order to calculate missing image data. Images processed this way typically suffer from blur, but even unexpected sharpness and colour banding can occur depending on the implementation and the image it's processing.
This method is most common as it is fast to calculate and usually looks pretty good from regular viewing distances. You can think of it as image stretching.
Bicubic Interpolation looks at groups of 16 pixels (4x4) in order to calculate the final image. This is great for colour accuracy, but no so great for sharpness. However, if a piece of video hardware goes to the trouble of using Bicubic Interpolation, they'll probably have their own post-processing effects like sharpening and noise reduction to get the best possible result. Our example uses those post processing techniques to balance colour accuracy, sharpness and blur as much as possible.
Because of the computational power needed to process real-time video in this manner, not all devices go to this much trouble when upscaling.
As you can see above, upscaling can't find new image data to increase the actual resolution of the source image. A video stream from a DVD only has 720x576 pixels of image information to to give to an upscaling device, so all the other pixels on a 1080p TV have to be calculated somehow (that's 1.65 million missing pixels!)
Here's what upscaling can't do:
This example is a native resolution image at double the size of the source (think of it as comparing a DVD image to a Blu-ray image). Here we can see all the data that upscaling can't see, and can't guess. It's sharper, cleaner and ultimately superior. We'd like to run a native resolution all the time, but not all content is always available in a source format that matches our TVs and Projectors. With 4K screens on the market, upscaling is even more important - a DVD on a 4K screen will never look as good as a Blu-ray on a 4K screen, and that won't be as good as actual 4K content. If you can, avoid upscaling for the best possible picture quality.
As you can see, not all upscaling methods are equal, and it's not always possible to get native resolution content. Chances are that lower-priced TVs and Projectors will use basic stretching (Bilinear interpolation), but that doesn't mean you're stuck with their default output. Look for options in the image menus of your display for Sharpness and/or Noise Reduction settings. One or both should be available on your TV or Projector.
These two go hand in hand to try and improve the clarity of the final image by using post processing effects not otherwise supported by the source device. Applying these effects too much can lead to colour banding, loss of detail or haloing, so go easy and try with different films. Testing with a black and white movie from the early 1930s will make a modern blockbuster look very artificial!
If all else fails, you may need to scour the market for a more powerful upscaling device to sit between your media player and your display. Ones that use Bicubic interpolation and post-processing effects in a manner that is appreciably superior to basic Bilinear + Sharpness methods may be expensive and hard to come by, so temper any advertising claims of image wizardry against the actual limitations of pixel data as shown above - "Zoom and Enhance" only works on episodes of CSI and Star Trek!
If you have any unanswered questions about Upscaling, please don't hesitate to contact us!